Image-guided adaptive radiotherapy is an emerging paradigm intended to deal with the anatomical changes that a patient undergoes during treatment. These changes can include deformation and relative movement of organs, tumor shrinkage, tissue swelling, and other phenomena that alter the way radiation interacts with the anatomy. The changes can occur on a timescale of days, hours, or seconds. By adapting the original treatment plan to these changes, image-guided radiotherapy has the potential to provide optimal delineation and targeting of tumors and critical structures at any moment during treatment. However, this requires acquiring and working with a large volume of patient imaging data. The scientific objective of this project is to investigate novel methods to condense sequences of CT imaging data into a unified temporal representation of the patient's anatomy as it changes during the treatment process. The practical goal is to create a suite of image processing tools that will enable the routine application of image-guided adaptive radiotherapy techniques in the clinic. The health benefit will be more precise dose targeting, enabling dose intensification and improving tumor control. Daily adaptive therapy requires rapid, minimally-supervised recalculation and transfer of treatment planning data between images of non-rigid anatomy. The first aim of this project is to develop a fast and automatic deformable image registration process for this purpose. In its second aim, the project will use these deformable registration concepts and tools to investigate novel methods to reconstruct time-dependent CT images by matching deformation models to sequences of planar image projections. This will enable one to monitor anatomical change under conditions where conventional 3D imaging is impractical and at the same time will enable a reduction in the supplemental radiation dose from imaging. The third aim of the project is to develop measurement and validation procedures to assess the accuracy and reliability of these novel image registration and reconstruction techniques. [unreadable] [unreadable] [unreadable]